Friday, September 30, 2005

The Shining: Feel good movie of the year?

Someone recut a trailer for Stanley Kubrick's "The Shining" to make it seem like a feel good movie about a man named Jack, who's a failed writer looking for inspiration and a boy, Danny, who's looking for a father. It uses every sleazy movie trailer trick in the book complete with Peter Gabriel feel good soundtrack. Outstanding!

Saturday, September 24, 2005

Tech Museum and Microsoft 30th Anniversary

To celebrate their 30th anniversary, Microsoft sponsored free admission to the San Jose Tech Museum. Since I hadn't been to the Tech Museum since it opened in 2000, we decided to check it out.

Wednesday, September 21, 2005

How to make a mosaic from Post - It notes

What do you do if your bored and have a ton of multi-colored Post-it notes? Make a mosaic! Via Grady Booch

Tuesday, September 20, 2005

REVIEW: The (Mis)behavior of Markets

I recently read The (Mis)Behaviour of Markets by Benoit Mandelbrot, Richard L. Hudson - the former discovered Fractal Geometry. Mandelbrot's thesis is that the world of financial modeling is based on some questionable concepts and that better models can be made by using fractal geometry. Many of the ideas in the book are similar to those that appear in Nassim Taleb's Fooled by Randomness (Taleb and Mandelbrot have co-authored some papers on the topic). The first half of the book is a convincing attack on the premises of modern finance by focusing on points of chaos in the history of the stock market (the 1987 crash, the 1997 Russian Crisis). The standard model of finance is based on a Gaussian distribution of price changes - a bell curve with a large number of small price changes and a small number of large changes. Most of the time this model works but extrodinary events break the model. Unfortunately, the extrodinary events matter a lot - if you remove the 10 days with the largest price changes in the S&P 500 the current value would be 2x higher. Mandelbrot goes on the discuss different distribution models such as the Cauchy distribution on which the market might be based. He then goes on to dispute another pillar of financial analysis - that yesterdays price change has no effect on today. He makes a convincing case that, at a minimum, price volatility is correlated - big changes yesterday make big changes today more likely. The net effect of this is that the standard financial models underestimate risk and the likelihood that there will be large price corrections. He also attacks the Capital Asset Pricing Model, the Efficient Market Hypothesis and the Black - Scholes Equation which are all based on the same financial underpinnings. The second portion of the book is somewhat weaker. Mandelbrot develops his own financial model based on fractal geometry and fractal time. Like Wolfram's "A New Kind of Science" there is a tendency to generate pretty graphs and then argue that it looks true therefore it is true. To be fair, this book is aimed at a general audience and delving into the analytic and computational process to vet his fractal model is probably beyond the scope. Mandelbrot does do some interesting statistical analysis of the price change distribution to show the failings of the standard model. Needless to say, the fractal model matches the observed data far better. Mandelbrot makes some effort throughout the book to convince people that this is not a "get rich" book - if anything the message is that the market is far riskier than you think. The book made for an entertaining read and did a good job of explaining the current state (and failings) of financial theory.

Flickr Fiesta

We went to the Flickr Fiesta at the Yahoo! Sunnyvale campus last night - some photos here. Imagine a party where everyone has a camera. Overall it was fun although I felt a bit quesy after scarfing down cotton candy during Jerry Yang's speech :-(

Sunday, September 18, 2005

Cereality: Two Spoons Up!

As promised, I visited the all-ceral cafe Cereality which had recently opened in Chicago's Loop across from the mercantile exchange. We went late in the evening on a week night so the place was empty. I ordered the "You Snooze, You Lose" which was Cinnimon Golden Grahams, Espresso and chocolate - a healty dinner indeed. The decor is an homage to cereal culture so we flopped in some poofy chairs and watched cartoons while we ate. The pictures tell the tale.

A new name

I changed the name of this blog for no reason in particular - unless boredom is a reason. Fear not, I still love the film Full Metal Jacket.

Wait Wait Don't Tell Me

On a recent trip to Chicago, my wife and I attended a taping of the NPR Quiz Show Wait Wait Don't Tell Me. It's really more of a comedy show with panelists and callers answering questions on the absurdity of the news - the prize is getting Carl Castle's voice on yout anwering machine. I've listened to this show for a while so I jumped at the chance to attend a live taping at the Bank One Auditorium. The taping lasts 2.5 hours on a Thursday night but the show is only an hour on the radio - some of this is due to needing to re-record sections for audio quality and technical glitches. But there is a lot more of the show that just needs to be cut to make the show into an hour. Personally, I wouldn't mind if the show was longer with less polish based on the live taping. After the show, the host and panelists answer questions from the audience which are mostly trivia about the show. For example, someone asked Carl Castle if he would record anything on people's answering machine messages and he said that he draws the line at blatant commercials (and apparently he's been asked). If your in Chicago and have the time, I highly recommend it. The only disappointment was that the guest was NPRs Terry Gross but she was doing the show remotely. The week before they had had Barack Obama on the show and he was live. Also, Richard Roeper (from Ebert & Roeper at the Movies) was supposed to be a panelist but couldn't make it so I guess we will have to attend another taping the next time we are in Chicago.

Thursday, September 15, 2005

JWZ on Theology

By way of reviewing The Exorcism of Emily Rose Jamie Zawinski analyzes the theological underpinnings of the film:
But man, religious people are weird. It always seems like they've heard of Occam's Razor, but they just don't quite get how it works! They kept saying things like, "God allows people to be posessed to prove to others that God exists". Well you know what, if God really wanted to prove that he existed, I don't think he'd have any trouble doing that, being God and all. Instead of making a statue bleed in front of some backwoods hick, why not make ten thousand statues bleed at the same time? It's fuckin' God! So the obvious, clich├ęd answer to that is that God doesn't actually want to provide proof, because he wants people to have faith (AKA "believing something for no reason at all"). In which case, posession proves nothing except that, well, God's kinda mean. In fact, providing proof of God would be more up the Devil's alley, wouldn't it? Proof would destroy faith. So is God skulking around like the Men In Black covering up Satan's spoilers? Also there was some nonsense about 3AM being "the witching hour" because Jesus came back from the dead at 3PM. Which immediately made me ask, what time zone is God in? And does he follow Daylight Savings Time?

Friday, September 09, 2005

WolframTones: Cellular Automata for Music

Wolfram Research (the makers of Mathematica) have released a service that uses cellular automata combined with music theory to generate unique musical compositions. They sound ... unique but they do sound better and more structured than a random combination of sounds. The ideas of using cellular automata as generators for large scale structure (similar to fractals) was popularized in Stephen Wolfram's book "A New Kind of Science".

Wolfram also runs the indispensible Mathworld reference site.

Saturday, September 03, 2005

Malcolm Gladwell on Open Source Cookies

This weeks New Yorker (Sept 5, 2005 - it's not online but sooner or later it will show up here) has an article by Malcolm Gladwell called "The Bakeoff" about the quest to create the ultimate cookie. The article profiles Mattson & Co, a Burlingame company that does independent development of new food products for most of the large food corporations and it's leader, Steve Gundrum. Gundrum wanted to develop a cookie that was healty (or at least healthier) than current cookies but that didn't sacrifice anything in taste. Gundrum apparently ran into Steve Jobs at an Apple Store (aside: I know about 5 people who this has happened to - does Jobs just hang out after work and sell Macs to people one at a time?) and met Mitch Kapor at a conference that turned him onto Open Source design ideas. He figured, designing software was like designing cookies recipies and some of the same ideas might be applicable. Gundrum has also heard of extreme programming (XP) so he decided to stage a contest between competing ideologies to develop the ultimate cookie.

  • One team was a standard Mattson product development team
  • The second team consisted of two development people working together doing rapid iteration a la eXtreme Programming - kind of. Rapid iteration is one of several requirements of XP - for example, it's not clear how unit testing applies when you're developing a cookie...
  • The third team (referred to as "The Dream Team") consisted of a group of Mattson employees that were being directed by a mailing list of 15 industry food development experts from a variety of companies in a pseudo Open Source model.

The teams ran into different sort of problems. The Open Source Dream Team spent a lot of time bickering:

Eventually, Carol Borda, the Dream Team project leader, asked Gundrum whether she should try and calm things down. He told her no; the group had to find its "own kind of natural rhythm." He wanted to know what fifeteen high-powered bakers thrown together on a project felt like, and the answer was that the felt like chaos. They took twice as long as the XP team. They created ten times the headache. Worse, no one in the Open Source team seemed to be having any fun. ... To Dan Fletcher, of Kellog's, "the whole thing spun in place for a long time. I got frustrated with that. The number of people involved seemed unweidly. You want some diversity of youth and experience, but you want to keep it close-knit as well. You get some depth in the process versus breadth. We were a mile wide and an inch deep"
The article goes on to highlight the weaknesses of the open source development model (citing Joel Spolsky) and the oft made argument that open source excels only at cloning existing products like web browsers and Unix. Ok, perhaps but the cookie experiment is not exactly conclusive evidence that Open Source doesn't work because it doesn't follow the model very closely. I think they would have had much better results (and been closer to the open source model) if they had had the 15 experts go off and develop their own cookies independently, then see which ones were good and have the experts iteratively tinker with the best cookies. The net effects of Open Source development may be social but at the work level, it's individual people doing individual tasks.

The result of all this was three different cookies which were mailed to 300 households with the following results (I have no doubt numerous critical details were ommitted about how the study was conducted - this is the New Yorker, not JAMA):

  • 14% of the respondents preferred the XP Developed Cookie
  • 41% of the respondents preferred the Open Source Developed Cookie
  • 44% of the respondents preferred the Mattson Development Team Cookie
Gladwell says the results are "close but unequivocal" - I'd say it's too close to make any real conclusion but as I said, there isn't enough info about how the study was done to really know. Apparently it was unequivocal because the Mattson cookie (its described as a "strawberry cobbler") is being put into development soon. More interestingly, Mattson apparently will license the cookie recipe at no cost to manufacturers provided a percentage of the profits go to a charitable organization - perhaps they did learn something from the open source model after all.

The article is definitely worth reading but (as a software person) I found the description of how foods are developed more interesting than their attempts to adapt software design methodology. I've written about Gladwell's excellent work in the past:

UPDATE: Jon Udell comments on the article here